By John Hejduk and Peter Wagner
The Census Bureau counts Wisconsin prisoners as if they were residents of the communities where they are incarcerated, even though they can't vote and remain legal residents of the places they lived prior to incarceration.
Crediting thousands of mostly urban and minority men to other communities has staggering implications for modern American democracy, which uses the Census to apportion political power on the basis of equally-sized state and county legislative districts.
Until recently, it didn't matter where the Census Bureau counted prisoners. As recently as 1980, the Wisconsin prison system had less than 4,000 prisoners, compared to 5 times as many in 2000. Using population as the sole measure of political apportionment in state and local government is also quite new. Counting prisoners as residents of the facility might have made sense in the country's first century, when the mandate of the Census was limited to determining the relative populations of states. But today, state legislatures use the Census to apportion political power within each state.
But Wisconsin, like most states, says that prison is not a residence. A legal residence is the place that people choose to be and do not intend to leave. The statutes are explicit: when people leave home temporarily with an intent to return, they retain their former residence. Since prisoners do not choose the prison and will return home the minute they are released from custody, their legal residence remains at their pre-incarceration address. While the Census Bureau counts most people at their legal residences, its method of counting prisoners is in fundamental opposition to how most states define residence.
Today, our conception of democracy requires far more detailed and accurate data to reflect where the population -- including people in prison -- actually lives. In almost every respect, except for how it counts people in prison, the Census Bureau's methodology has evolved to keep pace with the changing needs for its data. But now that such a large percentage of the population is incarcerated, where the Census Bureau counts people in prison is a question of critical importance.
Counting incarcerated people as if they were residents of prison towns leads to misleading portrayals of such communities.
Wisconsin has the second highest Black incarceration rate in the country, and the fifth highest racial disparity in incarceration, with Blacks 10.6 times as likely to be in prison as Whites. Counties with large prisons, though, tend to be disproportionately White: 87% of the state and federal prison cells are located in counties that are have a larger White population than the state as a whole. In Dodge County, 89% and in Marquette County, 91%, of the Black population reported in the Census is not residents, but prisoners.
The prison communities also tend to be small enough that incarcerated populations are a significant portion of the total "residents" counted by the Census. Twenty-four percent of the population reported in the Census for Waupun City (in Dodge and Fond du Lac Counties) is actually prisoners at the Waupun, Dodge and John C. Burke Correctional Facilities. About 5% of the "residents" counted in Dodge and Jackson Counties are actually prisoners. In Marquette County, more than 8% of "residents" are incarcerated.
There is also a geographic disparity in who goes to prison in Wisconsin. The residents of Milwaukee, Racine, Kenosha and Rock counties are much more likely to be incarcerated than the residents of other counties. The residents of Milwaukee County are more than twice as likely to be in prison than the average resident of the state, and more than 7 times as likely as the residents of prison-hosting Dodge County. Milwaukee County contains 18% of the state population and is home for 42% of its prisoners.
The Census Bureau's practice of counting prisoners as residents of the prison location complicates using the Census for demographic analysis of rural communities, but this problem is overshadowed by the serious damage the prisoner miscount does to state and local democracy.
The basic principle of American representative democracy is that every vote must be of equal weight. When governments draw districts with equal populations, they ensure that each resident has equal access to government, no matter where she or he lives. When districts are of substantially different sizes, the weight of each vote starts to differ: in underpopulated districts, each vote is worth more, and in overpopulated districts, a vote is worth less.
The U.S. Supreme Court first declared that the "One Person, One Vote" principle applied to state legislative redistricting in the 1963 landmark case Reynolds v. Sims. The Court struck down an apportionment scheme for the Alabama state legislature that was based on counties and not population. In 1960 Alabama, Lowndes County, with 15,417 people, had the same number of state senators as Jefferson County, with 634,864 people, giving the residents of sparsely-populated Lowndes County 41 times as much political power as the residents of densely-populated Jefferson County. The Supreme Court ruled that the 14th Amendment's equal protection clause required that districts be drawn to be substantially equal in population.
Subsequent U.S. Supreme Court cases defined the limits of "substantially equal." In White v. Regester, the Court ruled that the State of Texas was not required to justify how it drew lines resulting in an average district deviation of less than 2% and a maximum deviation of 9.9%. Today, most states draw their districts so that the smallest district is no more than 5% smaller, and the largest no more than 5% larger, than the average district. This keeps the difference between the largest and smallest district within 10%.
Wisconsin has historically applied a much higher standard, drawing districts with a maximum deviation of less than 2%. Only four states currently have districts that are more equal in population than Wisconsin's. For three decades, federal judges have drawn the state Assembly and Senate legislative district maps. In 1982, at the first redistricting since the U.S. Supreme Court allowed Texas to have a population deviation of 10%, the federal judges who drew Wisconsin's districts set a higher standard, explaining that "We believe that a constitutionally acceptable plan should not deviate as high as 10%, and should, if possible, be kept below 2%." The plan they drafted met even that high standard: "The deviation in our plan is a scant 1.74%."
In 1992, the court drew a plan with an even smaller total deviation from exact population equality: 0.52%. In 2002, the court drew a plan with a deviation of only 1.48%, still within the 2% threshold established in 1982.
Wisconsin rightly prioritizes population equality when drawing districts, but the Census Bureau has undermined these efforts by crediting thousands of prisoners to the wrong place.
In Wisconsin, a State Assembly district is supposed to contain 54,179 people, and -- according to the Census -- each district is close to that size. But the federal judges who drew the districts relied on Census data which counted people in prison in the wrong place.
The population in Assembly District 42, represented by J. A. Hines (R), includes 2,052 prisoners as 3.8% of its population. Of those, the 1,277 prisoners in the Federal Correctional Institution at Oxford are not even necessarily from the state of Wisconsin, or even physically located in the district. The Census Bureau mistakenly counted the prison in Oxford even though it is actually 12 miles away in Adams County.
District 53, represented by Carol Owens (R), is even more distorted by the Census methodology: it contains 5,131 prisoners from other parts of the state who were counted as residents of District 53. The actual population of the district is 48,905, 9.7% smaller than it should be. Using Census counts of prisoners to pad out the district gives each group of 9 residents in District 53 as much political power as 10 residents elsewhere in the state.
According to the Census, District 53 is 4.3% Black, a larger percentage than 81 other districts. But 90% of the 2,342 Blacks in the district are actually incarcerated people from other parts of the state.
Using inaccurate data to draw legislative districts creates a vote dilution problem in Wisconsin that is slightly larger than that allowed by the U.S. Supreme Court  and 5 times larger than that allowed by the federal judges who have for decades drawn Wisconsin's state legislative districts.
The most affected districts are listed in Table 1, and information on all Assembly districts is listed in Appendix A.
|District||Representative (2008)||Prisoners (state and federal)||District population||Population Deviation||Percent of district that is incarcerated||Percent deviation without prisoners|
The Census methodology also distorts Senate districts. Almost 4% of District 18, represented by Carol Roessler (R), is the 6,110 prisoners at facilities in Waupun, Oshkosh and Fond du Lac. See Appendix B for a table of prison populations in all Senate districts.
Counting prisoners as residents of the voting districts in which they are incarcerated does serious damage to county government and rural democracy. The fundamental inaccuracies of the data create a difficult situation for county officials. If counties rely on the data, the districts will be drawn unfairly. Dodge and Jackson are two counties that took two different approaches to this situation, but they are by no means the only Wisconsin counties facing considerable difficulty because of the Census methodology.
In Dodge County, each district is supposed to contain 2,208 people. Dodge County relied on Census data during redistricting to draw properly-sized districts. But the county contains three large prisons, each of which holds more than 1,200 people. Because the Census numbers included prison populations, more than half of the population in three districts is made up of prisoners. As a result, every resident in districts 8, 29 and 31 has twice the political power of residents of other districts.
Jackson County drew complicated districts that split a prison into four parts to alleviate the problem, but even this was insufficient to avoid serious vote dilution. A district in Jackson County is supposed to contain 1,005 people, but the Jackson Correctional Institution contains 953 people. As the County Clerk explained, a district with so few eligible voters is unlikely to produce any candidates, so the county attempted to reduce the prison's impact by dividing it equally between four districts. The prison is so large, however, that 24% of the population of each of these districts is prisoners. Every group of 76 residents in these 4 districts has as much of a say over county affairs as 100 residents elsewhere.
Many other counties and small cities with prisons in Wisconsin are similarly affected. Table 2 summarizes the impact of relying on Census data in 12 Wisconsin counties and cities.
|County District||Percent of district's population that is in prison||Resulting Vote Distortion|
|Adams 5 & 6||64%||9 votes here = 25 elsewhere|
|Brown 14||22%||39 votes here = 50 elsewhere|
|Dane 33||6%||47 votes here = 50 elsewhere|
|Dodge 29||53%||47 votes here = 100 elsewhere|
|Dodge 31||59%||41 votes here = 100 elsewhere|
|Dodge 35||10%||9 votes here = 10 elsewhere|
|Dodge 8||54%||23 votes here = 50 elsewhere|
|Fond du Lac 36||37%||63 votes here = 100 elsewhere|
|Jackson 10, 11, 12 and 19||24%||19 votes here = 25 elsewhere|
|Racine 13||17%||83 votes here = 100 elsewhere|
|Sheboygan 22||6%||47 votes here = 50 elsewhere|
|Sheboygan 32||25%||3 votes here = 4 elsewhere|
|Winnebago 12||42%||58 votes here = 100 elsewhere|
|Winnebago 30||16%||21 votes here = 25 elsewhere|
|Fitchburg City 4||14%||43 votes here = 50 elsewhere|
|Franklin City 1||38%||31 votes here = 50 elsewhere|
|Waupun City 2||63%||37 votes here = 100 elsewhere|
|Waupun City 3||79%||21 votes here = 100 elsewhere|
The Census Bureau's decision to count people in prison as if they were residents of the prison creates a serious vote dilution problem in Wisconsin cities and counties with prisons. Since even the counties that draw complicated districts to divide the prisons have been unable to eliminate the problem, it is clear that the data must be fixed before redistricting begins.
The Census Bureau's decision to credit thousands of disenfranchised non-residents to the Census blocks with prisons creates serious problems for Wisconsin. This problem need not exist in the future, as the state can lobby the Census Bureau for change, or fix the data itself.
The U.S. Supreme Court requires states to draw new districts each decade on the basis of population, but states are not required to use the Census. The Census Bureau, which collects its data at great cost, wants states to use its data and has historically been responsive to the needs of its data users when deciding how to count the population.
The ideal place to fix the prisoner miscount is at the U.S. Census Bureau. The method of counting other special populations has changed numerous times, in each case responding to changing demographics and needs. For example, when more college students began studying far from home, the Census policy changed in order to more accurately reflect American living situations. The Census Bureau considers state and local redistricting to be the second most important use of its data and should therefore be responsive to requests for a new way of counting prisoners.
Wisconsin state and local officials should lobby the Bureau directly and endorse the recent letter sent by New York State Senator Eric Schneiderman and St. Lawrence County New York Legislator Tedra Cobb. The letter asks the Census Bureau to count prisoners as residents of addresses outside the facility, and if insufficient time remains before the 2010 Census to implement that change, that the Bureau publish a special version of the PL94-171 redistricting data file for the prison population. Such detailed block-level counts of prison populations would assist the state and its counties in identifying and removing the prison populations prior to redistricting.
While states must redistrict on the basis of actual population, the U.S. Constitution does not require states to use the federal census for its own redistricting. Wisconsin can fix the Census data by creating a special state-level census that collects the home addresses of people in prison and then adjusts the U.S. Census counts prior to redistricting. Legislation with these goals is currently pending in New York, Illinois, and Michigan, and is patterned on how Kansas adjusts the federal census counts of military and students.
Many Wisconsin cities and counties with prisons drew districts based on the Census and thus diluted the votes of residents who do not live next to the prisons. While many of these cities and counties felt obligated to use the Census when drawing the districts, they may use other sources of data or correct the Census data to remove the prison population.
Mississippi, Colorado and New Jersey require counties with prisons to remove the prison population prior to redistricting, and Virginia law encourages it. Many other counties decide to exclude the prison population prior to redistricting. In Essex County, New York, the county Board of Supervisors enacted a law that mandated excluding the prisons when apportioning its government because "the inclusion of these federal and state correctional facility inmates unfairly dilutes the votes or voting weight of persons residing in other towns within Essex County."
In Michigan, nearly all counties avoided distorting democracy by ignoring the prisoners in drawing the districts, whether the potential for distortion was very large or quite small. Gratiot County modified data to avoid creating a district that would have been 50% prisoners. In Lapeer, using the census would have meant a district with 6% prisoners, but even there the county clerk told us that they excluded prisoners because the prisoners were "not really residents."
If the U.S. Census and the state of Wisconsin fail to change how prisoners are counted in the next Census, counties and cities with prisons should remove the prison populations from their redistricting data prior to redistricting. This would ensure that each resident is given the same access to local government, regardless of whether she or he lives next to a large prison.
This report uses the correctional facility populations as counted by the Census Bureau and published in Summary File 1, Table P37. Our analysis is based on a subset of this data, filtered to include only state or federal prisoners and not people in jail, halfway house residents, etc. Our filtering methodology is documented in Section 2 of the Democracy Toolkit (2007). The Census does not publish race and ethnicity for the group quarters population at the tract level, so we exported our data to Arcview 9.2, where we manually deduced the race and ethnicity of each facility's population from tract-level data made available in tables PCT17B and PCT17H. In the few cases where this was insufficient, we accessed race and ethnicity data for the individual blocks from Tables P7 and P11, which were then attributed to the correctional facility.
As discussed in footnote 8 above, the Census placed the Federal Correctional Institution at Oxford in the wrong county. The facility's mailing address is in Marquette County, but the prison is actually in Adams County. This mistake was corrected as a part of the Count Question Resolution program and was correctly reflected in the county level redistricting, but the correction did not arrive in time to be used in the state legislative redistricting process.
We believe that the Census Bureau double-counted the Green Bay Correctional Institution in Brown County, as the Census reported 1,936 in the facility but our historical facility population counts suggest that the figure should have been about 1,018 people. However, as this is the figure reported by the Census, and because we can identify no other population -- incarcerated or otherwise -- that should be attributed this population, our analysis uses the same data that the legislature did: the Census's count of 1,936 prisoners.
Finally, the Census Bureau misclassified the Robert E. Ellsworth Correctional Center as part of the "other nonhousehold living situations" at the Southern Wisconsin Center for the Developmentally Disabled, on whose grounds the prison sits. Because this was a correctional population counted in the Census and this data was used to draw legislative districts, our analysis counted this population as a correctional population. However, this classification mistake made it impossible to deduce the race and ethnicity of the people incarcerated at the Robert E. Ellsworth Correctional Center, so we used the counts taken on June 30 2000, by the Bureau of Justice Statistics, Census of State and Federal Correctional Facilities, 2000: 251 people incarcerated, of which 90 were Black and 70 were Latino.
The impact on individual state districts and counties was calculated in ArcView 9.2 based on shapefiles distributed by the Census Bureau and the Wisconsin Legislative Redistricting Information website.
Our analysis of the impact on county districts follows the methodology in the Democracy Toolkit to determine whether prison populations were excluded from districts and to calculate the resulting vote dilution in counties that included them. Most counties told us that they felt obligated to include the prison populations in the districts, and we set out to calculate the impact on their democracy that resulted. Most counties did not publish the population totals for their districts, so we calculated the population totals by overlaying Census data over the county district maps in Arcview. In the counties that publish electronic files with the shapes of their districts, this calculation was straightforward. In other counties, we recreated their paper maps in Arcview and then calculated the population totals using both official Census data and a special version of the Census data with the prison populations removed. In all cases, the total population (including the prison population) in each district was approximately the ideal district size for that county, confirming that the districts were indeed based on prison populations.
The incarceration rates for counties, published in Appendix D, were calculated from Professor Pamela Oliver's Excel spreadsheets for Wisconsin county imprisonment rates data [XLS], and the official Census 2000 population counts for counties, to represent the rates at which county residents are sent to prison. (Federal and local jail imprisonment would therefore be excluded from these relative comparisons.)
This research was supported by a grant from the The After Prison Initiative of the Open Society Institute. We thank Michael Keane of the Wisconsin Legislative Reference Bureau and Tony Van Der Wielen of the Legislative Technology Services Bureau for speaking with us about how Census data and the Count Question Resolution program corrections were handled in the Wisconsin redistricting process and the many county clerks who generously gave their time to help us understand how their county government was structured. Pamela Oliver, Dan Jenkins and Bill Cooper gave very helpful feedback of an earlier draft of this report and Will Goldberg helped to improve the text.
John Hejduk is a law clerk at the Prison Policy Initiative and a second-year student at the Western New England College School of Law.
Peter Wagner is an attorney and Executive Director of the Prison Policy Initiative. In 2002, he authored the first district-by-district analysis of the impact of Census counts of prisoners on state legislative redistricting, Importing Constituents: Prisoners and Political Clout in New York (2002). He has presented his research at national and international conferences and meetings, including a Census Bureau Symposium, a meeting of the National Academies, and keynote addresses at Harvard and Brown Universities. His publications include, with Rose Heyer, Too Big to Ignore: How Counting People in Prisons Distorted Census 2000 (2004) and, with Eric Lotke, Prisoners of the Census: Electoral and Financial Consequences of Counting Prisoners Where They Go, Not Where They Come From [PDF] (2005).
The Massachusetts-based non-profit, non-partisan Prison Policy Initiative documents the impact of mass incarceration on individuals, communities, and the national welfare. We produce accessible and innovative research to empower the public to participate in creating better criminal justice policy.
 Ann L. Pastore and Kathleen Maquire eds., Sourcebook of Criminal Justice Statistics 2003, Table 6.29, 501 (2005) and U.S. Census 1980 and 2000.
 "The residence of a person is the place where the persons habitation is fixed, without any present intent to move, and to which, when absent, the person intends to return." Wis. Stat. § 6.10 at P 1 (2002).
 "A person shall not lose residence when the person leaves home and goes into another state or county, town, village or ward of this state for temporary purposes with an intent to return." Wis. Stat. § 6.10 at P 5 (2002).
 Mark Mauer and Ryan S. King, Uneven Justice: State Rates of Incarceration By Race and Ethnicity Table 3, 8 (July 2007).
 Mark Mauer and Ryan S. King, Uneven Justice: State Rates of Incarceration By Race and Ethnicity Table 6, 11 (July 2007).
 Peter Wagner, Outdated methodology impairs Census Bureau's count of Black population, Prisoners of the Census Fact of the Week for May 3, 2004.
 Calculated from Professor Pamela Oliver's Excel spreadsheets for Wisconsin county imprisonment rates data [XLS], and the U.S. Census 2000.
 This last figure is doubly ironic because the prison facility isn't even in Marquette. The Census mistakenly counted Oxford Prison at its postal address in Marquette County, when it actually is in neighboring Adams County. (The published Census figures were used for state redistricting and remain the official numbers for most purposes, but the distortion would be almost the same if the prison had been counted in the right place. The Census Count Question Resolution program states that Adams County has 19,920 people, of whom 6.4% are incarcerated.)
 Reynolds v. Sims, 377 US 533 (1964).
 White v. Regester, 412 U.S. 755 (1973).
 Illinois, California, Washington and Minnesota drew more equal House districts, and Illinois, California, Florida and Washington drew more equal Senate districts. National Conference Of State Legislators, Redistricting 2000 Population Deviation Table, (visited August 8th, 2007)
 AFL-CIO v. Elections Bd., 543 F. Supp. at 634 (E.D. Wis. 1982).
 AFL-CIO at 637.
 Prosser v. Elections Bd., 793 F.Supp. 859, 870 (W.D. Wis. 1992).
 Baumgart v. Wendelberger, (2002). Wendelberger Slip Copy, 2002 WL 34127471 at 8. Also available at http://www.legis.state.wi.us/ltsb/redistricting/Decision/decision.pdf
 White v. Regester, 412 U.S. 755 (1973).
 We also looked at Oshkosh City in Winnebago Co., Fond du Lac City in Fond Du Lac Co., Allouez Village in Brown Co., and Sturtevant Village in Racine Co., but these cities and villages are not affected because their local government is elected at large rather than from districts. Marquette County's districts were not affected by the prisoner miscount because the census error that mistakenly placed the Federal Oxford prison in Marquette was corrected in time for county redistricting.
 In Bethel Park v. Stans, 449 F. 2d 575 (C.A. 3, 1971) the city of Philadelphia and a Pennsylvania congressman sued the Census Bureau for counting military personnel, students and prisoners at their temporary addresses instead of their home addresses, because the plaintiffs feared a loss of representation. The court stated that "Although a state is entitled to the number of representatives in the House of Representatives as determined by the federal census, it is not required to use these census figures as a basis for apportioning its own legislature." Bethel Park v. Stans, 449 F. 2d 575, 583 (C.A. 3, 1971). Two years later, when faced with a similar issue, the Supreme Court rejected Virginia's argument that it was compelled to use the Census Bureau's assignment of residences to military personnel when drawing state legislative districts. Mahan v. Howell, 410 U.S. 315, 331 (1973).
 New York Senate Bill 2754 is the ideal version of this legislation as it includes a specific provision that would apply the census adjustment not only to senate and assembly districts, but to county districts as well. The bill was first introduced by Senator Schneiderman on February 25th, 2005.
 Illinois House Bill 7338 was introduced by Rep. Arthur Turner on October 22nd, 2004.
 Michigan House Bill 4395 was introduced by Reps. Lemmons, Young and Gonzales on June 19th, 2007.
 In Kansas, "Senatorial and representative districts shall be reapportioned upon the basis of the population of the state adjusted: (1) To exclude nonresident military personnel stationed within the state and nonresident students attending colleges and universities within the state; and (2) to include military personnel stationed within the state who are residents of the state and students attending colleges and universities within the state who are residents of the state in the district of their permanent residence." KS CONST. art. 10 § 1.
 Baraga, Branch, Chippewa, Gratiot, Ionia, Jackson, Lapeer, Lenawee, Manistee, Marquette, Montcalm, and Muskegon Counties all draw county legislative districts without prisoners. Luce County last updated its districts in 1992, before the prison was built. Washtenaw and Saginaw Counties drew legislative districts that contain the prison populations.